Article Predictive Biomarker for Progression Into the Sunset Glow Fundus of Vogt-Koyanagi-Harada Disease, Using Adaptive Binarization of Fundus Photographs

Yuki Komuku1, Hiroto Ishikawa1,AtsuyaIde2,TaikiMatsuoka1, Hisashi Fukuyama1, Takeshi Okadome2, and Fumi Gomi1

1 Department of Ophthalmology, Hyogo College of Medicine, Hyogo, Japan 2 Kwansei Gakuin University, Hyogo, Japan

Correspondence: Yuki Komuku, Purpose: The sunset glow fundus (SGF) appearance in Vogt-Koyanagi-Harada (VKH) Department of Ophthalmology, disease was evaluated by means of adaptive binarization of patients’ fundus Hyogo College of Medicine, 1-1 photographs. Mukogawa-cho, Nishinomiya, Methods: Hyogo, 663-8501, Japan. e-mail: Twenty-nine Japanese patients with acute VKH were enrolled in this study. We [email protected] evaluated one eye of each patient, and thereby divided the patients into two groups; SGF+ and SGF− at 6 months after treatment. We compared patient age, gender, and Received: July 8, 2020 spherical equivalent refractive error (SERE) and choroidal thickness measured using Accepted: September 5, 2020 optical coherence tomography. We also compared the choroidal vascular appearance Published: October 9, 2020 index (CVAI), derived by adaptive binarization image processing of fundus photographs, Keywords: Vogt-Koyanagi-Harada between the two groups. Measurements of choroidal thickness and CVAI were taken at disease; sunset glow fundus; the onset of disease, and 1, 3, and 6 months after treatment. The sunset glow index (SGI), adaptive binarization; uveitis as previously reported, was calculated using color fundus photographs, and compared to the CVAI. Citation: Komuku Y, Ishikawa H, Ide A, Matsuoka T, Fukuyama H, Results: Eight patients (27.6%) were categorized into the SGF+ group. At all time points, Okadome T, Gomi F. Predictive the mean CVAI in the SGF+ group was significantly greater than that in the− SGF group. biomarker for progression into the No significant difference was observed in choroidal thicknesses at any time point.The sunset glow fundus of SGI was significantly greater in the SGF+ group at 6 months. Vogt-Koyanagi-Harada disease, Conclusions: using adaptive binarization of CVAI could be a new predictive biomarker for the development of SGF in fundus photographs. Trans Vis Sci patients with VKH disease. Tech. 2020;9(11):10, Translational Relevance: Detecting SGF is important for management of patients with https://doi.org/10.1167/tvst.9.11.10 VKH, and CVAI may indicate the possibility of developing into SGF, although the color fundus photographs do not yet show SGF at that time.

was reported by Suzuki, and is an objective evaluation Introduction of the color balance in color fundus photographs.5 The second measure is polarization-sensitive optical Vogt-Koyanagi-Harada (VKH) is a systemic coherence tomography (PS-OCT), which was reported autoimmune disorder that affects organs, includ- by Miura et al., and can provide an in vivo objec- ing the eyes, that have melanocytes in their tissues.1 tive evaluation of choroidal melanin loss underlying In some patients with VKH, progression into chronic the SGF in chronic VKH disease.6 Unfortunately, recurrences of granulomatous uveitis occurs because PS-OCT is not a widely used form of optical coherence depigmentation of the choroid results from choroidal tomography (OCT) and thus it is not an option at melanocyte damage.2,3 In those cases, the fundus many facilities. shows orange discoloration called “sunset glow fundus Although ocular inflammation, progressing subclin- (SGF)”.1,4 There are few objective measures of SGF. ically in chronic VKH disease with SGF, is sight- The first one is the sunset glow index (SGI), which threatening, detection of such progression is still

Copyright 2020 The Authors tvst.arvojournals.org | ISSN: 2164-2591 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 2 challenging. For example, Keino et al. reported that Measurement of CVAI 67.5% of patients showed SGF, although chronic clini- cal ocular inflammation was apparent in only 17.5% of As introduced above, the CVAI is the choroidal the patients.7 Murata et al. have reported that signs of vasculature appearance index of the binarization image 1 = subclinical ocular inflammation were detected in most from color fundus pictures. Briefly, CVAI CVR/IR × patients with VKH with SGF by indocyanine green 1000, where CVR is the number of pixels of the angiography.8 As an alternative, early post-treatment choroidal blood vessel region and IR is the number of choroidal thickness was reported to be an index to alert pixels in the imaging region. CVAIwas calculated using physicians to future progression to SGF.9 original software and we confirmed the repeatability Recently, we defined the choroidal vascular appear- of the software results. A total of 116 color fundus ance index (CVAI) as the visibility of choroidal photographs and 116 OCT images were evaluated. vessels in color fundus photographs, based on adaptive We used color fundus photographs (TRC-50DX, binarization imaging processing.10 Our previous study Topcon, Tokyo, Japan) with similar brightness, indicated that the CVAI was correlated with choroidal contrast, and color balance characteristics. Choroidal thickness. In this study, we investigated whether the thickness under the fovea was measured with the CVAI adds specific information to the SGF, like the enhanced depth-imaging technique on an OCT device 12,13 SGI does, and whether the progression of SGF can be (Spectralis, Heidelberg, Germany). Choroidal predicted by evaluating the CVAI. thickness that could not be measured because the The aim of this study was to investigate how inner scleral border was not visible on OCT (over CVAI changes with choroidal thickness, and how SGF 800 μm) was set to 800 μm. As a result, the choroidal progresses because of decreasing melanin in patients thickness was set to 800 μm in 19 images (14.7%). with VKH disease, to determine whether CVAI could be a predictive biomarker for the development of SGF. Measurement of the Sunset Glow Index To compare the color balance of the fundus photographs, we calculated the SGI, as in the previ- Methods ous studies.5,6,14 The formula is as follows: SGI = Lred /(Lred + Lgreen + Lblue), where Lred,Lgreen,and Study Design and Eligibility Lblue represent the mean luminance of the red, green, and blue (R, G, and B) channels, respectively. Between January 2016 and May 2019, a total of 29 Japanese patients with acute VKH were enrolled in this retrospective, observational study. We excluded Study End Points recurrent cases, patients suspected of having a recur- rent case with SGF at the initial visit, and cases with The primary end point was the difference in the strong anterior chamber inflammation. The current CVAI between the fundus pictures from patients with and without SGF. All eyes were classified into one of study was performed in accordance with the Decla- + − ration of Helsinki and with approval from the ethics two groups, SGF and SGF , based on the fundus committees of Hyogo College of Medicine (No. 2426). photographs at 6 months post-treatment. Photographs of each eye were individually evaluated by two blinded observers (Y.K. and H.F.) regarding the presence Study Protocol of SGF. When the two observers disagreed, a third (H.I.) judged the photograph and the three observers The data we extracted from medical records together reached a consensus. Comparisons of CVAI included patient age, gender, and spherical equivalent with choroidal thickness and SGI, between the two refractive error (SERE). Measurements of choroidal groups at each time point, were secondary end points. thickness, SGI, and the CVAI were taken at the onset We also assessed the baseline characteristics of all of disease, and 1, 3, and 6 months after treatment. The patients. diagnosis of VKH was made in accordance with the revised criteria proposed by the International Nomen- Statistical Analyses clature Committee.11 We used the data for the right eye of each patient. No patients underwent intraocular For continuous variables, the mean, median, and surgery. All patients were treated with systemic corti- range were calculated. For discrete variables, the costeroids from the disease onset, starting with steroid number of values in each category and the percentages pulse therapy. in each category were calculated. Group differences

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 3 Table. Baseline Characteristics of Patients With or Without Sunset Glow Fundus (SGF) SGF− Versus Total SGF− Group (n = 21) SGF+ Group (n = 8) SGF+ P Value Age, y 50.7 ± 18.8 54.3 ± 17.6 41.3 ± 19.7 0.09 SERE, D −1.13 ± 3.24 −0.69 ± 2.55 −2.28 ± 4.61 0.24 Choroidal thickness, μm 693 ± 130 698 ± 132 679 ± 133 0.74 CVAI, % 0.041 ± 0.022 0.040 ± 0.024 0.046 ± 0.019 0.549 SGI 0.59 ± 0.048 0.59 ± 0.050 0.59 ± 0.040 0.89 CVAI, choroidal vascular appearance index; SERE, spherical equivalent refractive error; SGI, sunset glow index.

were assessed using the Student’s t-test or Wilcoxon characteristics, age, SERE, choroidal thickness, CVAI, signed-rank test for continuous variables and Fisher’s and SGI in all eyes for the SGF− group and SGF+ extracttestorthePearsonχ 2 test for categorical group are shown in the Table. Briefly, no significant variables. differences in any parameters were observed between We assessed the accuracy of each screening test the SGF− and SGF+ groups at baseline. using receiver operating characteristic (ROC) curve analysis. ROC analysis provides several advantages over sensitivity and specificity determination for a Primary End Point: CVAI in Patients With and single cutoff. The formula for calculation of the Without SGF “true positive rate” was true positive/(true positive + The mean CVAI was significantly greater at 1, 3, false negative); the formula for calculation of the and 6 months than at baseline in both of the SGF− “false positive rate” was false positive/(false positive + < = + and SGF groups (P 0.0001, P 0.0028, and true negative). Calculating the area under the ROC P = 0.0047 in the SGF− group, and P < 0.0001, curve (AUC) provides a summary of discriminative P < 0.0001, and P = 0.0026 in the SGF+ group, respec- ability for a screening test and allows quick compar- tively). The mean CVAI values in the SGF+ group ison of discriminative ability among different screen- at 1, 3, and 6 months were significantly greater than ing tests. The AUC has a value from 0.5 to 1.0, those in the SGF− group (P = 0.072, P < 0.0001, and where 1.0 represents perfect ability to discriminate P < 0.0001, respectively, Fig. 1A). between, for example, children without vision disorders Typical color fundus photographs, the correspond- and children with vision disorders, and 0.5 represents 15 ing adaptive binarization images, basic picture for discrimination resulting from pure chance. An AUC CVAI, histograms of pixel values for each channel of greater than 0.9 is considered excellent; greater than 0.8 the RGB image, raw data for SGI, and OCT images through 0.9 is very good; 0.7 to 0.8 is good; 0.6 to 0.7 is 16 are shown in Figure 2. The extent of the whitish area average, and less than 0.6 is considered poor. Analy- in the adaptive binarization images from the SGF+ ses were performed with JMP Pro (version 14.0.0; SAS group tended to exceed that in adaptive binarization Institute Inc., Cary, NC). For all analyses, P values were images from the SGF− group at 1, 3, and 6 months. reported, as well as two-sided 95% confidence intervals SGI showed no change during the follow-up period in for point estimates. Statistical significance was deter- = < the SGF– group, but significantly increasedP ( 0.038) mined when P values were 0.05. at 6 months in the SGF + group. In the histograms, each channel of the RGB image (excluding the frame part) is shown. OCT images in both groups showed Results rapid resolution of the retinal detachments after the therapy. Baseline Characteristics Secondary End Points Altogether, we used 116 fundus photographs and OCT images from 29 patients with VKH (mean age of Choroidal Thickness, SGI, and SERE With and Without 50.7 ± 18.8 years, range = 13–81 years). All patients SGF had characteristic symptoms of VKH in both eyes. The mean choroidal thickness was significantly Of the 29 patients, 8 (27.6%) showed SGF. Baseline thinner at 1 (353 μm), 3 (352 μm), and 6 months

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 4

Figure 1. Changes in the choroidal vascular index (CVAI), choroidal thickness, sunset glow index (SGI) and spherical equivalent refractive error (SERE) in the SGF− and SGF+ groups. (A) The mean CVAI was significantly elevated at all post-treatment examinations in both groups, compared with baseline. The mean CVAI in the SGF+ group at 1, 3, and 6 months was significantly greater than that in the SGF− group (P = 0.072, P < 0.0001, and P < 0.0001, respectively). (B) The mean choroidal thickness decreased significantly from baseline across time in both groups; however, no significant differences were found between the− SGF and SGF+ groups. (C) The mean SGI did not change significantly from baseline at 1, ,3 or 6 months in −the SGF group. However, it was significantly greater at 6 months than at baseline in the SGF+ group and was significantly greater than that in the SGF− group at that time. (D) The mean SERE did not change significantly from baseline at any point in either the SGF− or SGF+ groups. Additionally, no significant difference was found between the groups at any point.

(359 μm) post-treatment than at baseline (697 μm; group was significantly greater than that (0.59) in the P < 0.0001 for each time) in the SGF− group, and at SGF− group (P = 0.027; Fig. 1C). 1 (424 μm), 3 (306 μm), and 6 months (320 μm) post- The mean SERE did not change significantly treatment than at baseline (679 μm) in the SGF+ group from baseline at any time point in either the SGF− (P < 0.0001, P < 0.0001, and P = 0.0018, respectively). or SGF+ group, and no significant difference No significant difference was found between the− SGF was observed between the two groups at any time and SGF+ groups at any point (baseline, 1, 3, and (Fig. 1D). 6 months; P = 0.74, P = 0.30, P = 0.33, and P = 0.42, respectively, Fig. 1B). The mean SGI did not differ significantly between ROC Curves and AUCs Between CVAI and SGI its baseline (0.590) and the values at 1 month (0.580), The AUCs for CVAI ROC curves were 0.78 at 3 months (0.588), and 6 months (0.589) in the SGF− 1 month, 0.94 at 3 months, and 0.95 at 6 months. group (P = 0.50, P = 0.90, and P = 0.97, respectively). For the SGI, the corresponding AUCs were 0.55 at However, the SGI was significantly greater at 6 months 1 month, 0.59 at 3 months, and 0.74 at 6 months, (0.636) than at baseline (0.587) in the SGF+ group indicating that the CVAI was more sensitive at all time (P = 0.038). No significant difference at baseline, or points. 1 and 3 months was found between the SGF− and Furthermore, on the basis of the Youden index, the SGF+ groups (P = 0.85, P = 0.84, and P = 0.39, cutoff value for CVAI that later developed into SGF respectively), but that at 6 months (0.64) in the SGF+ was 0.0808 at 1 month and 0.0818 at 3 months (Fig. 3).

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 5 reported that SGF occurs because, in addition to the loss of melanocytes, tissue damage after inflamma- tion can result in tissue necrosis, fibrosis, and, finally, choroidal thinning.3,17 However, in the present study, no significant difference in choroidal thickness was found between the SGF− and SGF+ groups. Hirooka et al. reported that the mean central choroidal thickness at 1 week after the start of treatment was significantly greater in eyes with SGF than in eyes without SGF, and that the cutoff choroidal thickness value at 1 week (to signal future progression to SGF) was 410 μm.20 In the present study, although no significant difference was found, the mean choroid thickness in the SGF+ group was greater at 1 month (424 vs. 353 μm) than in the SGF− group, and smaller at 3 months (306 vs. 352 μm) and 6 months (319 vs. 359 μm). These results suggest that spontaneous inflammation and thick choroid in the early phase after the start of treatment may be a risk factor for progression to SGF. Therefore, earlier detec- tion of the SGF or earlier prediction of progression to SGF could lead to a reduction in chronic inflammation, Figure 2. Typical color fundus photographs and corresponding by promoting early treatments. adaptive binarization images, histograms of pixel values and optical Regarding the diagnosis of SGF, according to a coherence tomography (OCT) images. Color fundus photographs previous study, the mean SGI was 0.649 in SGF+ and and corresponding adaptive binarization images, histograms of − 6 pixel values and OCT images from an SGF− patient (upper) and an 0.578 in SGF . Here, the mean SGI was 0.636 in the SGF+ patient (lower). There was no change in appearance across SGF+ group and 0.589 in the SGF−. Furthermore, the time in the adaptive binarization images from the SGF− patient. CVAI differed earlier between the SGF− and SGF+ However, the whitish area in the adaptive binarization images from groups than did the SGI. ROC curve analyses and the SGF+ patient tended to increase during the follow-up period. AUCs indicated that the CVAI was a good biomarker The histograms show each channel of the RGB image (excluding at 1 month and an excellent biomarker at 3 and the frame part), and the x- and y-axes are drawn on the same scale 6 months, whereas SGI was poor at 1 and 3 months, used in the other panels. In the OCT images, subretinal detachment greatly improved after therapy. and only a good biomarker at 6 months. Therefore, as a predictive biomarker, CVAI is more sensitive than SGI in the early stage of VKH disease. In the evaluation of CVAI, each color fundus photograph was separated Discussion into 8-bit RBG (red, green, blue) components, and the R-component image was used in the detection of This study demonstrated that the CVAI was greater choroidal vessels. The R component was also used for in the SGF+ group than in the SGF− group at 1, 3, and SGI calculation, however, not to evaluate the choroidal 6 months after treatment, despite no significant differ- vessels but rather the redness of the color fundus ence in choroidal thickness being found between the photographs, as determined by both the choroidal and SGF− and SGF+ groups at those time points, and no retinal vessels. Therefore, we propose that during the significant difference in SGI at 1 and 3 months. There- development of the SGF, choroidal vascular-related fore, the CVAI could be the most sensitive marker to changes occur before the fundus becomes reddish. The detect SGF in the early phase of VKH disease. orange-red discoloration of SGF is due to depigmen- The appearance of SGF is associated with thinning tation of the inflamed choroid,18 and the choroidal of the choroid during the convalescent phase of VKH stroma contains a large number of melanocytes with disease.17 In the convalescent phase, depigmentation melanin.21 Miura et al. provided objective evalu- of the choroids occurs 2 to 6 months after disease ation of melanin loss in chronic VKH disease by onset.18 The convalescent stage with SGF was report- using PS-OCT, and eyes with SGF showed a lack of edly characterized by mild-to-moderate inflammatory choroidal melanin and the preservation of melanin cell infiltration.4 Fong et al. suggested that stromal in the retinal pigment epithelium.6 Thus, the decrease scarring causes shrinkage and dropout of small vessels in melanin content may increase the transparency of in the choroid in the convalescent stage.19 Nakai et al. choroidal vessels. We suggest that in the process of

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 6

Figure 3. Receiver operating characteristic (ROC) curves and areas under the curve (AUCs) obtained from CVAI and sunset glow index (SGI) in patients with or without sunset glow fundus (SGF). ROC curves and AUC values based on the choroidal vasculature appearance index (CVAI) and sunset glow index (SGI) in patients with or without SGF at 1 month (left), 3 months (middle), and 6 months (right) after treatment. The red dots show the cutoff values.

decreasing choroidal melanin, the medium and large significant myopic progression developed severe SGF choroidal vessels are the first to become more visible, within 4.3 ± 3.1 years, and the mean refractive error and second that redness develops because of further was −3.17 diopter (D) in progression of myopia in thinning of the choroidal stroma and severe decrease > −1.0 D and −2.89 D in no progression myopia group in melanin, causing the CVAI to differ between the and no significant difference was found.24 Addition- SGF− and SGF+ groups before SGI differences ally, no significant difference was observed in our short occur. Furthermore, of 8 patients with SGF, only follow-up study, thus we think that a smaller SERE 1 developed the SGF at 3 months after treatment, may be a risk factor leading to SGF, and that myopia but the other 7 patients progressed to the SGF by progresses over a long interval. 6 months. As described above, CVAI in SGF+ patients This study had several limitations. First, it was a already tended to be greater than in SGF− patients retrospective study and conducted at a single center, by 1 or 3 months. Therefore, CVAI may indicate the with a relatively small number of patients and the possibility of developing into SGF, although the color number of SGF+ patients was particularly small (all fundus pictures do not yet show any SGF at that time. of whom were Japanese). Second, the broad range of Moreover, CVAI changed more over time than did SGI. ages (13–81 years old) of the included patients was a CVAI was approximately 2.6 times greater at 3 months limitation because choroidal thickness is strongly influ- (0.12 ± 0.026) and 2.5 times greater at 6 months enced by age. Third, in this study, adaptive binariza- (0.11 ± 0.029) than it was at baseline (0.046 ± 0.019), tion was used for image processing. Unlike the normal whereas SGI was only approximately 10% greater at binarization, adaptive binarization determines whether 6 months (0.64 ± 0.055) than at baseline a certain pixel is assigned a 0 or 1 based on periph- (0.59 ± 0.040). eral pixel values. Normally, choroidal blood vessels The results showed that the mean SERE was smaller have a higher pixel value than the retina, thus it in the SGF+ group than in the SGF−, although no was possible to extract the choroidal vessels only by significant difference was found. It has been reported adaptive binarization. However, when the pixel values that the choroid thins with aging and with myopic became higher overall, due to progression of the SGF, degeneration,22,23 because choroidal thickness was choroidal vessels may not be detected accurately by significantly correlated with the degree of tessella- adaptive binarization. Therefore, a long-term observa- tion,14 and thinning of the choroid is associated with tional study is needed to evaluate the progression of the progression to SGF.17 It was reported that eyes with SGF.

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 7 In conclusion, CVAI could be a novel predictive 10. Komuku Y, Ide A, Fukuyama H, et al. Choroidal biomarker to indicate the development of SGF in thickness estimation from colour fundus pho- patients with VKH disease. tographs by adaptive binarisation and deep learn- ing, according to central serous chorioretinopathy status. Sci Rep. 2020;10:5640. 11. Read RW, Holland GN, Rao NA, et al. Revised Acknowledgments diagnostic criteria for Vogt-Koyanagi-Harada dis- ease: Report of an international committee on The authors thank Claire Barnes, PhD, from nomenclature. Am J Ophthalmol. 2001;131:647– Edanz Group (https://en-author-services.edanzgroup. 652. com/ac) for editing a draft of this manuscript. 12. Spaide RF, Koizumi H, Pozonni MC, Pozonni MC. Enhanced depth imaging spectral-domain Disclosure: Y. Komuku, None; H. Ishikawa, None; optical coherence tomography. Am J Ophthalmol. A. Ide, None; T. Matsuoka, None; H. Fukuyama, 2008;146:496–500. None; T. Okadome, None; F. Gomi, None 13. Margolis R, Spaide RF. A pilot study of enhanced depth imaging optical coherence tomography of the choroid in normal eyes. Am J Ophthalmol. 2009;147:811–815. References 14. Yoshihara N, Yamashita T, Ohno-Matsui K, Sakamoto T. Objective analyses of tessellated 1. O’Keefe GAD, Rao NA. Vogt-Koyanagi-Harada fundi and significant correlation between degree of disease. Surv Ophthalmol. 2017;62:1–25. tessellation and choroidal thickness in healthy eyes. 2. Rao NA. Pathology of Vogt-Koyanagi-Harada PLoS One. 2014;9(7):e103586. disease. Int Ophthalmol. 2007;27:81–85. 15. Hanley JA, McNeil BJ. The meaning and use of 3. Inomata H, Sakamoto T. Immunohistochem- the area under a receiver operating characteristic ical studies of Vogt-Koyanagi-Harada dis- (ROC) curve. Radiology. 1982;143:29–36. ease with sunset sky fundus. Curr Eye Res. 16. Choi BCK. Slopes of a receiver operating charac- 1990;9(Suppl):35–40. teristic curve and likelihood ratios for a diagnostic 4. Read RW, Rechodouni A, Butani N, et al. Compli- test. Am J Epidemiol. 1998;148:1127–1132. cations and prognostic factors in Vogt-Koyanagi- 17. Nakai K, Gomi F, Ikuno Y, et al. Choroidal obser- Harada disease. Am J Ophthalmol. 2001;131:599– vations in Vogt-Koyanagi-Harada disease using 606. high-penetration optical coherence tomography. 5. Suzuki S. Quantitative evaluation of “sunset glow” Graefes Arch Clin Exp Ophthalmol. 2012;250:1089– fundus in Vogt-Koyanagi-Harada Disease. Jpn J 1095. Ophthalmol. 1999;43:327–333. 18. Goto H, Rao NA. Sympathetic ophthalmia and 6. Miura M, Makita S, Yasuno Y, et al. Polarization- Vogt-Koyanagi-Harada syndrome. Int Ophthalmol sensitive optical coherence tomographic documen- Clin. 1990;30:279–285. tation of choroidal melanin loss in chronic Vogt- 19. Fong AH, Li KK, Wong D. Choroidal eval- Koyanagi-Harada disease. Investig Ophthalmol Vis uation using enhanced depth imaging spectral- Sci. 2017;58:4467–4476. domain optical coherence tomography in Vogt- 7. Keino H, Goto H, Usui M. Sunset glow fundus Koyanagi-Harada disease. Retina. 2011;31(3):502– in Vogt-Koyanagi-Harada disease with or without 509. chronic ocular inflammation. Graefe’s Arch Clin 20. Hirooka K, Saito W, Namba K, et al. Early post- Exp Ophthalmol. 2002;240:878–882. treatment choroidal thickness to alert sunset glow 8. Murata T, Sako N, Takayama K, et al. Iden- fundus in patients with Vogt-Koyanagi-Harada tification of underlying inflammation in Vogt- disease treated with systemic corticosteroids. PLoS Koyanagi-Harada disease with sunset glow One. 2017;12(2):e0172612. fundus by multiple analyses. J Ophthalmol. 21. Weiter JJ, Delori FC, Wing GL, Fitch KA. Reti- 2019;2019:3853794. nal pigment epithelial lipofuscin and melanin and 9. Hirooka K, Saito W, Namba K, et al. Early post- choroidal melanin in human eyes. Investig Ophthal- treatment choroidal thickness to alert sunset glow mol Vis Sci. 1986;27:145–152. fundus in patients with Vogt-Koyanagi-Harada 22. Ding X, Li J, Zeng J, et al. Choroidal thickness disease treated with systemic corticosteroids. Wal- in healthy Chinese subjects. Investig Opthalmology lace GR, ed. PLoS One. 2017;12:e0172612. Vis Sci. 2011;52:9555.

Downloaded from tvst.arvojournals.org on 10/03/2021 Predictive Biomarker for the SGF of VKH Disease TVST | October 2020 | Vol. 9 | No. 11 | Article 10 | 8 23. Harb E, Hyman L, Gwiazda J, et al. Choroidal 24. Takahashi H, Takase H, Terada Y, Mochizuki thickness profiles in myopic eyes of young adults M, Ohno-Matsui K. Acquired myopia in Vogt– in the correction of myopia evaluation trial cohort. Koyanagi–Harada disease. Int Ophthalmol. Am J Ophthalmol. 2015;160:62–71.e2. 2019;39:521–531.

Downloaded from tvst.arvojournals.org on 10/03/2021